Gartner says AI coding costs will pass the average developer’s salary by 2028. Forbes ran that prediction this week, next to Uber burning through its whole AI budget in four months and $1.7 million of billing mistakes found inside $34 million of token spend.
Before we panic, what actually has to be true for that prediction to land?
First, consumption has to keep surging even as the price per token falls. It will. That’s the old Jevons paradox… make a resource cheaper and we don’t use less of it, we find more uses for it. Agentic development alone could add an order of magnitude.
Second, and this is the part the headline skips: that spend has to buy less than the salary it’s compared to. Nobody knows that yet. A token bill bigger than a salary is a bargain if it ships what three engineers used to. It’s a bonfire if it doesn’t. The bill alone can’t tell you which, right?
If it were my Monday morning, I’d take the biggest AI line item and put a denominator under it. Cost per merged change, per resolved ticket, per closed case. If nobody can name the denominator, that’s your finding.
Source: After ‘Tokenmaxxing’, Token Spend Has Become The New Metric To Watch (Forbes, Tim Keary, 2026-07-10)
Originally posted on LinkedIn.
RationalAI helps organizations turn ideas like this into results: AI strategy and advisory services. Previous article: Agent reliability rhymes with self-driving’s last 10%
